no. 335 march 4, 1999 cars, cholera, and cows · 2016-10-20 · no. 335 cars, cholera, and cows the...

No. 335

CARS, CHOLERA, AND COWSThe Management of Risk and Uncertainty

by John Adams

Executive Summary

Everyone takes and manages risks, balancing potentialrewards against uncertain losses. Experts provide adviceabout managing risk, but with limited success. Part oftheir difficulty stems from an incomplete appreciation ofthe different kinds of risk and part from their inadequateconsideration of the different mindsets with which peoplerespond to risks.

Directly perceptible risks are managed instinctivelyand intuitively. Professional attempts to manage them arethwarted by people who insist on being their own risk man-agers. Risks perceptible with the help of science includeinfectious diseases. Quantitative estimates of such risksare frustrated by the reflexivity of risk. Virtual risksinclude mad cow disease and suspected carcinogens.Scientists do not know or cannot agree about the nature ormagnitude of these risks, and nonscientists argue frompreestablished beliefs, convictions, and superstitions.Imaginary or not, virtual risks have real consequences forindividuals, corporations, and governments.

This paper examines the cultural filters through whichpeople view virtual risks and describes some of the long-standing debates about virtual risks that arise because ofthose filters. Understanding our ignorance of the risksand filters is a precondition for civilized debate aboutvirtual risks. ____________________________________________________________

John Adams is professor of geography, University College,London, England. He is the author of Risk (1995) and "Vir-tual Risk and the Management of Uncertainty," in the RoyalSociety's Science, Policy and Risk (1997).

March 4, 1999

Introduction

Every day, in decisions ranging from crossing thestreet to considering whether a hamburger is safe to eat,every one of us must face and manage a wide variety ofrisks. Every such decision involves balancing the uncer-tain rewards of actions against the potential losses.Figure 1 uses the risk "thermostat" as a model of thisbalancing act. The model postulates that

• everyone has a propensity to take risks;

• this propensity varies from one individual toanother;

• this propensity is influenced by the potentialrewards of risk taking;

• perceptions of risk are influenced by experience ofaccident losses--one's own and others';

• individual risk-taking decisions represent a balanc-ing act in which perceptions of risk are weighedagainst propensity to take risks; and

• accident losses are, by definition, a consequenceof taking risks; the more risks an individual takes,

Figure 1The Risk "Thermostat"

the greater, on average, will be both the rewards andthe losses he incurs.

There has been a long-running and sometimes acrimo-nious debate between "hard" scientists--who treat risk assomething that can be objectively measured--and social sci-entists--who argue that risk is culturally constructed.Much of this debate has been caused by the failure of theparticipants to distinguish among different kinds of risk.It is helpful, when considering how the balancing act inFigure 1 is performed, to be clear about the sort of riskone is dealing with. There are three:

• directly perceptible risks, such as climbing atree, riding a bicycle, and driving a car;

• risks perceptible with the help of science, such ascholera and other infectious diseases; and

• virtual risks, about which scientists do not orcannot agree, e.g., the connection between bovinespongiform encephalopathy (BSE, or mad cow disease)and CJD (Creutzfeldt-Jakob disease) in humans; globalwarming; and numerous suspected carcinogens.

In Figure 2 these categories are represented by threeoverlapping circles to indicate that the boundaries betweenthem are indistinct, and also to indicate the potentialcomplementarity of approaches to risk management that havepreviously been seen as adversaries in the debate betweenthe "hard" scientists and the cultural constructionists.

Figure 2Three Types of Risk

Perceivedthroughscience

Perceiveddirectly

Virtualrisk


Perceiveddirectly

Virtualrisk


Perceiveddirectly

Virtualrisk

Directly Perceptible Risks

Figure 1 can serve as a description of the behaviorof the driver of a single car going around a bend in theroad. His speed will be influenced by his perception ofthe rewards of risk; these might range from getting to thechurch on time to impressing his friends with his skill orcourage. His speed will also be influenced by his percep-tion of the danger; his fears might range from death,through the cost of repairs and loss of his license, tomere embarrassment. His speed will also depend on hisjudgment about the road conditions--is there ice or oil onthe road? How sharp is the bend and how high thecamber?--and the capability of his car--how good are thebrakes, suspension, steering, and tires?

Overestimating the capability of the car or the speedat which the bend can be safely negotiated can lead to anaccident. Underestimating those things will reduce therewards gained. The consequences, in either direction,can range from the trivial to the catastrophic.

The balancing act described by this illustration isanalogous to the behavior of a thermostatically controlledsystem. The setting of the thermostat varies from oneindividual to another, from one group to another, from oneculture to another, and for all of these, over time. Somelike it hot--a Hell's Angel or a Grand Prix racing driver,for example--others like it cool--a Caspar Milquetoast ora little old lady named Prudence. But no one wantsabsolute zero.1

Risk: An Interactive Phenomenon

Figure 3 introduces a second car to the road to makethe point that risk is usually an interactive phenomenon.One person's balancing behavior has consequences for oth-ers. On the road one motorist can impinge on another's"rewards" by getting in his way and slowing him down, orhelp him by giving way. One is also concerned to avoidhitting other motorists or being hit by them. Driving intraffic involves monitoring the behavior of other motor-ists, speculating about their intentions, and estimatingthe consequences of a misjudgment. Drivers who see a carapproaching at high speed and wandering from one side ofthe road to the other are likely to take evasive action,unless perhaps they place a very high value on their dig-nity and rights as a road user and fear a loss of esteemif they are seen giving way. During this interactionenormous amounts of information are processed.

Figure 3The Risk Thermostat: Two Drivers Interacting

Moment by moment each motorist acts upon informationreceived, thereby creating a new situation to which theother responds.

Figure 4 introduces a further complication. On theroad and in life generally, risky interaction frequentlytakes place on terms of gross inequality. The damage thata heavy truck can inflict on a cyclist or pedestrian is

Figure 4The Risk Thermostat: Truck Driver and Cyclist Interacting

great; the physical damage that a cyclist or pedestrianmight inflict on the truck is small. The truck driver inthis illustration can represent the controllers of largerisks of all sorts. Those who make the decisions thatdetermine the safety of consumer goods, working conditions,or large construction projects are, like the truck driver,usually personally well insulated from the consequences oftheir decisions. The consumers, workers, or users oftheir constructions, like the cyclist, are in a positionto suffer great harm, but not inflict it.

Risk Compensation: The Case of Seat Belt Legislation

The phenomenon illustrated by Figures 1, 3, and 4 iscommonly termed "risk compensation"2 or "offsetting behav-ior." As people perceive themselves as safer or betterequipped against a danger, they are more likely to takemore risks. Although the phenomenon is widely accepted astrue, it is almost universally denied in risk regulation.

The efficacy of seat belt legislation has become anaccepted "fact." Now that seat belt laws have been passedin more than 80 jurisdictions around the world, one wouldexpect the evidence in support of the claims for seat beltlegislation to be voluminous, but oddly it has shrunk dra-matically. The claims now all rest on the experience ofonly one country, the United Kingdom.3

In 1991, after surveying the global evidence, LeonardEvans of the General Motors Research Laboratory reachedthe following conclusion about seat belt legislation:

The highest precision evaluation is for the UK'slaw, where belt use rose rapidly from 40% to 90%in a large population of affected occupants.The law reduced fatalities to drivers and front-seat passengers by 20%. For smaller use rateincreases, and for smaller populations (that is,in nearly all other cases), it is not possibleto directly measure fatality changes. They canbe reliably estimated using an equation based onthe known when-used effectiveness of the beltstogether with a quantification of selectiverecruitment effect--the tendency of those chang-ing from nonuse to use to be safer than averagedrivers.4

In other words, according to Evans, of the more than80 jurisdictions with seat belt laws, fatality reductionscan be measured only in the United Kingdom. In all the

other jurisdictions the life-saving benefits were too smallto register in the accident statistics.

The claims made for seat belt laws in all otherjurisdictions rest on a deduction that assumes there is norisk-compensation effect. There is no basis for thatassumption. Indeed, there is a vast amount of evidence ofmeasurable response to interventions that influence theoutcomes of other sorts of "crashes"--trapeze artists withsafety nets, rock climbers with ropes, and lovers withcondoms, to cite three obvious examples. All attemptmaneuvers with their safety equipment that they would notattempt without it.

It is not clear why the proponents of seat belt leg-islation believe that protection in car crashes should bean exception to this well-established principle. In fact,elsewhere in his book, Evans indicates that potential out-come does influence behavior.

All drivers I have questioned admit that theywould drive more carefully if their vehicles con-tained high explosives set to detonate on impact;dramatically increasing the harm from a minorcrash can clearly reduce the probability of aminor crash.5

Evans's evidence concerning the life-saving benefitsof seat belts if one is in a crash is not disputed. It'sintuitively obvious that a person traveling at high speedinside a hard metal shell will stand a better chance ofsurviving a crash if he is restrained from rattling aboutinside, and there's an impressive body of empirical evi-dence showing that the use of a seat belt improves a caroccupant's chances of surviving a crash. Evans has calcu-lated that wearing a belt reduces one's chances of beingkilled if in a crash by 41 percent. He assumes that thisbenefit has been enjoyed by all those people in the 80-plus jurisdictions who belted up in response to a law, andthe laws therefore can be given credit for saving largenumbers of lives. But it does seem curious that with sucha large effect, and despite the fact that hundreds of mil-lions of motorists all around the world are now compelledby law to wear seat belts, he has no confidence in thedata to demonstrate directly measurable fatality reductionsexcept in the United Kingdom. Given the significanceattached to the United Kingdom result, we look at it moreclosely.

The UK Seat Belt Law

As a concession to the doubts that had been raisedabout seat belt efficacy in the early 1980s, Britain'sfirst seat belt law was passed for an unusual three-yeartrial period. It came into effect in January 1983 but wasnot made permanent until another vote in Parliament inJanuary 1986. During those three years, the Department ofTransport reduced the claim for lives saved from 1,000 ayear to 200.6 The lower figure was described as a "net"reduction; the decrease in the numbers of people killed inthe front seats of cars and vans in 1983 was partiallyoffset by an increase in the numbers of pedestrians,cyclists, and rear seat passengers killed.7 This shift infatalities was consistent with the risk-compensationhypothesis that predicted that the added sense of securityprovided by belts would encourage more heedless driving,putting other road users at greater risk. Despite thisimplicit acknowledgment of risk compensation, the evidenceon which Parliament relied when it confirmed the law in1986 was fundamentally flawed: it ignored the effect ofdrunken driving.

The number of road accident deaths had been decreas-ing at a near-steady rate from at least 1971 through 1981.In 1982, however, there was a noticeable increase that wasrestricted to the hours between 10:00 p.m. and 4:00 a.m.,the so-called drunk driving hours. The downward trend inroad accident deaths continued at all other hours in 1982.

The 1982 "alcohol blip" occurred almost entirely innon-built-up areas, and it has never been satisfactorilyexplained. In the words of a Transport and Road ResearchLaboratory Report, "The series for drinking car drivers innon-built-up areas shows an increase in 1982 which cannotbe related to available explanatory variables."8

Just as there was an increase in road accident deathsin 1982, there was a decrease in 1983 after the seat beltlaw went into effect. Although some experts jumped to theconclusion that the seat belt law accounted for thedecrease, that conclusion ignored the effects of the cam-paign initiated against drunk driving in 1983.

The decrease in fatalities in 1983 was clearly relat-ed to the campaign against drunken driving. In that year,

• "evidential" breath testing was introduced;

• unprecedented numbers of breath tests were adminis-tered;

• the number of motorists successfully prosecuted fordrunken driving increased by 31 percent;

• the decrease in road deaths between 10 at night and4 in the morning was 23 percent, while in all otherhours it was only 3 percent--in line with the pre-vailing trend; and

• the percentage of dead drivers whose blood alcoholwas over the legal limit dropped from 36 percent to31 percent.

In advocating the retention of the seat belt law inthe 1986 parliamentary debate, the Department of Transportrelied most heavily on the analysis of two statistics pro-fessors, James Durbin and Andrew Harvey from the LondonSchool of Economics. The time-series models developed byDurbin and Harvey for their analysis of the seat belteffect were impressively sophisticated, but none containedalcohol-related variables. They attributed all of thedecrease in fatalities in 1983 below the projected trendto the beneficial effect of the seat belt law, and none tothe campaign against drunken driving. In a presentationto a Royal Statistical Society seminar, Durbin and Harveyacknowledged that their analysis had taken no account ofalcohol and said that the effects of alcohol need futureresearch study. But no studies have so far explained whyseat belts have been so extraordinarily selective in sav-ing the lives only of those who are over the alcohol limitand driving between 10 at night and 4 in the morning.

In summary, there were two major road-safety measuresintroduced by the British government in 1983: the seatbelt law and the campaign against drinking and driving.In that year, there was a small, temporary drop in roadaccident fatalities below the established trend. The evi-dence with respect to seat belts suggests that the law hadno effect on total fatalities but was associated with aredistribution of danger from car occupants to pedestriansand cyclists. The evidence with respect to alcohol sug-gests that the decrease in fatalities in 1983 during thedrink-drive hours is accounted for partly by the still-unexplained rise above the trend in 1982 and partly by thedrink-drive campaign in 1983.

The evidence from Britain, which has been singled outas the only jurisdiction in the world in which it is pos-sible to measure fatality changes directly attributable toa seat belt law, suggests that the law produced no netsaving of lives. It did, however, redistribute the burdenof risk from those inside vehicles, who were already the

best protected, to those outside vehicles, who were themost vulnerable.

The Management of Directly Perceptible Risks

The management of directly perceptible risks--by toxi-cologists, doctors, the police, safety officials, andnumerous other "authorities"--is made difficult and frus-trating by individuals' insisting on being their own riskmanagers and overriding the judgments of risk experts andthe interventions of safety regulators--a phenomenon rou-tinely attested to by millions of smokers, sunbathers,consumers of jelly-filled donuts, and drinking and speedingmotorists. Why do so many people insist on taking morerisks than safety authorities think they should? It isunlikely that they are unaware of the dangers--there canbe few smokers who have not received health warnings, and,indeed, most smokers overestimate their risks. It is morelikely that the safety authorities are less appreciativeof the rewards of risk taking.

Directly perceptible risks are "managed" instinctive-ly; our ability to cope with them has been built into usby evolution--contemplation of animal behavior suggeststhat it has evolved in nonhuman species as well. Ourmethod of coping is also intuitive; we do not do a formalprobabilistic risk assessment before we cross the street.There is now abundant evidence, particularly with respectto directly perceived risks on the road, that risk compen-sation accompanies the introduction of safety measures thatdo not reduce people's propensity to take risks.Statistics for death by accident and violence, perhaps thebest available aggregate indicator of the way in whichsocieties cope with directly perceived risk, display astubborn resistance, over many decades, to the efforts ofsafety regulators to reduce them.9

Risk Perceived through Science

The risk and safety literature does not cover allthree categories of risk equally. It is overwhelminglydominated by the second category--risks perceived throughscience (Figure 5).

Central to this literature is the rational actor, acreature from the imaginations of risk experts,10 who man-ages risks on the basis of the experts' judgment about howa rational optimizer would, and should, act if in posses-sion of all relevant scientific information. In this

Figure 5The Dominance of the Rational Actor Paradigm in the Risk and Safety Literature

literature economists and scientists strive together toserve the interests of someone we might call Homo economi-cus-scientificus--the offspring of the ideal economist andthe ideal scientist.

Infectious diseases such as cholera are not directlyperceptible. One requires a microscope to see the agents

Figure 6Trends in Mortality: Britain 1841-1980

Source: British Medical Association, Living with Risk, 1987.

Perceiveddirectly Virtual

risk


0

50

100

150

200

250

300

350

400

1840 1860 1880 1900 1920 1940 1960 1980Sta

ndar

dize

d M

orta

lity

Rat

ios

(SM

R)

with

SM

R fo

r 19

50-1

952

set a

t 100

that cause them and scientific training to understand whatone is looking at. Science has an impressive record inmaking invisible or poorly understood dangers perceptible,and in providing guidance about how to avoid them.

Large, and continuing, decreases in premature mortali-ty over the past 150 years, such as those shown forBritain in Figure 6, have been experienced throughout thedeveloped world. Such trends indicate that ignorance isan important cause of death and that science, by reducingignorance, has saved many lives. When the connectionbetween balancing behavior and accidents that is shown inFigure 1 is not perceptible, there is no way that knowl-edge of cause and effect can inform behavior.

The Institutional Management of Risk

This is the realm of risk quantification. Everyindividual performs the mental balancing act described inFigure 1 in his or her own head. Institutions--governmentdepartments or large commercial enterprises--usually assignthe job of risk management to particular people or depart-ments who have no (or very little) balancing responsibili-ty and rarely consider rewards to be gained from particu-lar actions. Individuals confronting directly perceivablerisks usually make risk decisions informally and intuitive-ly. Institutions conduct the process explicitly and for-mally, wherever possible expressing risk in terms of mag-nitudes and probabilities.

Figure 7 sets out the sequence of steps recommendedin a formal risk assessment. Figure 8 describes a similarset of procedures used by a large pharmaceutical companyto manage risk.11 The risk literature is replete with sim-ilar algorithms. But, however sophisticated, the shadedoverlays that I have added show that when compared withFigure 1 they are simply more elaborate versions of thebottom loop of the risk thermostat model. Figures 7 and 8illuminate the propensity of risk managers to ignore therewards of risky behavior and the varying attitudes thatindividuals take toward risks.

Risk management in institutional settings, with a fewexceptions such as insurance and venture capital enterpris-es, turns out on inspection to be exclusively concernedwith risk reduction. Institutional risk management modelscharacteristically have no top loop; the "rewards" loop isthe responsibility of some other department, often market-ing. This view was reinforced during a seminar I present-

ed to the risk managers of a large private-sector concern,when one of the participants said, rather morosely inresponse to this suggestion, "Yeah, that's right. Aroundhere we're known as the sales prevention department." Thefollowing pronouncements from Shell Oil are typical ofinstitutional risk managers whose objective is the elimina-tion of all accidents.

The safety challenge we all face can be veryeasily defined--to eliminate all accidents thatcause death, injury, damage to the environment orproperty. Of course this is easy to state, butvery difficult to achieve. Nevertheless, thatdoes not mean that it should not be our aim, orthat it is an impossible target to aim for.12

The aim of avoiding all accidents is farfrom being a public relations puff. It is theonly responsible policy. Turning "gambling man"into "zero-risk man" (that is, one who managesand controls risks) is just one of the chal-lenges that has to be overcome along the way.13

Figure 7The Risk Assessment Process, Government Style

Source: A Guide to Risk Assessment and Risk Management for Environmental Protection (London: HerMajesty's Stationery Office, 1995).

Figure 8Risk Assessment Process in a Large Pharmaceutical Company

The single-minded pursuit of risk reduction by insti-tutional managers inevitably leaves the pursuers disap-pointed and frustrated. The risk thermostat model and theevidence supporting it suggest that safety interventionsthat do not lower the settings of the risk thermostats ofthe individuals at whom the interventions are aimed areroutinely offset by behavioral responses that reassert thelevels of risk that people were originally content with.This problem is compounded by the division of labor usual-ly found in institutional risk management; different peopleor departments are commonly placed in charge of the topand bottom loops--with no one obviously in charge of theoverall balancing act.

Some Limitations to Scientific Risk Management: A RichterScale for Risk?

Homo economicus-scientificus is an expert gambler,sensitive to small variations in the odds associated withthe risks he runs. The adherents to the rational actorparadigm, the authors of most of the "scientific" riskliterature, frequently express their dismay at the inabili-ty of ordinary people to make sensible use of such infor-mation. As one of their tasks, they seek ways to makelaypeople better informed and more "rational" in theirrisk-taking decisions. Central to the rationalist perspec-tive on risk management is the insistence that all riskscan, and should, be reduced to numbers.

In Great Britain the Department of Trade and Industryhas proposed the development of a "Richter Scale for Risk"that would "involve taking a series of common situationsof varying risk to which people can relate."14 The RoyalStatistical Society has called for "a simple measure ofrisk that [people] can use as a basis for decision mak-ing."15 The chief medical officer of health has called forthe development of an agreed standard scale for communicat-ing information about risk to the general public,16 and thecollection of risks presented in Table 1 is a typicalexample of what he had in mind.

The risk of dying in a road accident (1:8000) isintended to represent the average risk of dying in a roadaccident; it is derived by dividing the number of peoplekilled in a given year by the total population. It iscommonly found about halfway down such tables. It isincluded because road accidents are the most common causeof accidental death--and hence assumed to be a familiar"benchmark" risk to which people can relate for purposesof seeing other risks in proper perspective. But thereare a number of problems with this number that cast doubton the utility of the table as a guide to individual risk-taking decisions.

First, the number is out of date. The most recentnumber is 1:15686,17 about half the number in Table 1, mov-ing road accidents from the "low" to the "very low" cate-gory. But this error is trivial compared with the compli-cations that would arise should an individual seek to basea risk-taking decision upon it.

A trawl through the road safety literature18 revealsthat a young man is 100 times more likely to be involvedin a severe crash19 than is a middle-aged woman; someonedriving at 3 a.m. Sunday, 134 times more likely to diethan someone driving at 10 a.m. Sunday; someone with apersonality disorder 10 times more likely to die, andsomeone at two and a half times the blood alcohol limit 20times more likely to die. If these factors were all inde-pendent of each other, one could predict that a disturbed,drunken young man driving at 3 a.m. Sunday would be about2.7 million times more likely to be involved in a seriousroad accident than would a normal, sober, middle-agedwoman driving to church seven hours later.20

These four factors, of course, are not independent;there are almost certainly proportionately more drunken anddisturbed young men on the road in the early hours of themorning than at other times of day. But I have listedonly four complicating factors from a very long list.

Table 1Risk of an Individual's Dying (D) in Any One Year or Developing an Adverse Response (A)______________________________________________________________________________________Term Used Quantitative Risk Range Example Measured Risk______________________________________________________________________________________High Greater than 1:100 A. Transmission to susceptible 1:1-1:2

household contacts of measlesor chickenpox

A. Transmission of HIV from 1:6mother to child (Europe)

A. Gastrointestinal effects of 1:10-1:20antibiotics

Moderate 1:100-1:1000 D. Smoking 10 cigarettes per day 1:200D. All natural causes, age 40 years 1:850

Low 1:1000-1:10000 D. All kinds of violence and 1:3300poisoning

D. Influenza 1:5000D. Accident on road 1:8000

Very low 1:10000-1:100000 D. Leukemia 1:12000D. Playing soccer 1:25000D. Accident at home 1:26000D. Accident at work 1:43000D. Homicide 1:100000

Minimal 1:100000-1:1000000 D. Accident on railway 1:500000A. Vaccination-associated polio 1:1000000

Negligible Less than 1:10000000 D. Hit by lightning 1:10000000D. Release of radiation by nuclear 1:10000000

power station ______________________________________________________________________________________

Source: On the State of the Public Health: The Annual Report of the Chief Medical Officer of theDepartment of Health for the Year 1995 (London: Her Majesty's Stationery Office, 1996), p. 13.

Does the car have worn brakes, bald tires, a loose suspen-sion, and a valid registration? Is the road well lit,dry, foggy, straight, narrow, clear, congested? Does thedriver have good hearing and eyesight, a reliable heart, aclean license? Is the driver sleepy, angry, aggressive,on drugs? All these factors, and many more, can influencea motorist's chances of arriving safely.

A further complication is that the numbers citedabove relating to age, sex, time of day, and so forth arethemselves averages about which one would expect to findconsiderable variation. To the extent that the risks ofmotoring are felt to be directly perceptible, the risk-balancing behavior of motorists will be guided by theirindividual perceptions of risk and reward. So whether thenumber used for road accidents on the Richter Scale is

1:8000 or 1:16000, it is difficult to see how it couldserve as a guide to an individual risk-taking decision.

Consider another "familiar" risk comparison frequentlyfound in risk tables--the risk of death in an air crash.It is commonly asserted that the fear of flying is irra-tional, because "objectively" flying is safer than driving.John Durant, in a paper for the Royal Society's Conferenceon Science, Policy and Risk, sets out what might be calledthe orthodox-expert view of the safety of flying and theproblem created by popular "subjective biases."

The fact that many people behave as if theybelieve that driving a car is safer than flyingin an aeroplane (when on objective criteria theopposite is the case) has been attributed to acombination of the greater dread associated withplane crashes and the greater personal controlassociated with driving. Faced with a mismatchbetween scientific and lay assessments of therelative risks of driving and flying, few of usare inclined to credit the lay assessment withany particular validity. On the contrary we aremore likely to use the insight to help overcomeour own subjective biases in the interests of amore "objective" view.21

Evans succinctly deconstructs this view.22 He beginswith the most commonly quoted death rates for flying(0.6/billion miles) and road travel (24/billion miles) andcomes to a much less commonly quoted conclusion. He notes

1. that the airline figure includes only passengers,while the road figure includes pedestrians andcyclists;

2. that the relevant comparison to make with airtravel is the death rate on the rural Interstate sys-tem, which is much lower than the rate for the aver-age road;

3. that the average road accident death rates thatlead to the conclusion that it is safer to fly arestrongly influenced by the high rates of drunkenyoung men, while people dying in air crashes are, onaverage, much older and, when on the road, safer-than-average drivers; and

4. that because most crashes occur on take-off orlanding, the death rate for air travel increases astrip length decreases.

Taking all those factors into account he concludesthat a 40-year-old, belted, alcohol-free driver in a largecar is slightly less likely to be killed in 600 miles ofInterstate driving--the upper limit of the range overwhich driving is likely to be a realistic alternative toflying--than on a trip of the same distance on a scheduledairline. For a trip of 300 miles he calculates that theair travel fatality risk is about double the risk of driv-ing. This comparison, of course, is not the completestory. The risks associated with flying also should bedisaggregated by factors such as aircraft type and age;maintenance; airline; the pilot's age, health, and experi-ence; weather; and air traffic control systems.

Insurance

The insurance industry uses, generally successfully,past accident rates to estimate the probabilities associat-ed with future claim rates. This success is sometimesoffered as an argument for using the cost of insuringagainst a risk as a measure of risk that would be a usefulguide to individual risk takers. Weinberg has argued that"the assessment is presumably accurate, since in generalit is carried out by people whose livelihood depends ongetting their sums right."23

However, the fact that the livelihoods of those inthe insurance business depend on "getting their sumsright" does not ensure that the cost of insuring against arisk provides a good measure of risk for individuals. Thesum that the insurance business must get right is theaverage risk. For most of the risks listed in Table 1,the variation about the average will range, depending onparticular circumstances, over several orders of magnitude.Insurers depend on ignorance of this enormous variabilitybecause they need the good risks to subsidize the bad. Ifthe good and bad risks could be accurately identified, thegood ones would not consider it worthwhile to buy insur-ance, and the bad ones would not be able to afford it.

This is precisely the threat to the insurance busi-ness posed by discoveries about genetic predisposition tofatal illness. The greater the precision with which indi-vidual risks can be specified, the less scope remains fora profitable insurance industry. The current debate aboutwhether insurance companies should be allowed to demanddisclosure of the results of genetic tests focuses atten-tion on the threat to the industry of knowledge thatassists the disaggregation of these averages. If disclo-sure is not required, people who are poor risks will be

able to exploit the insurance companies; and if it isrequired, the insurance companies will be able to discrim-inate more effectively against the bad risks--making them,in many cases, uninsurable.

The companies want access to the results to protectthem from exploitation by customers who have access to theinformation. That knowledge converts risks to certainties(or at least permits risks to be specified more accurate-ly) and reduces the size of their market.

Problems of Measurement

Risk comes in many forms. In addition to economicrisks--such as those dealt with by the insurance busi-ness--there are physical risks and social risks, and innu-merable subdivisions of these categories: political risks,sexual risks, medical risks, career risks, artistic risks,military risks, motoring risks, legal risks. The list isas long as the number of adjectives that might be appliedto behavior in the face of uncertainty. These risks canbe combined or traded. People are tempted by the hazardpay available in some occupations. Some people, such assky-diving librarians, may have very safe occupations anddangerous hobbies. Some young male motorists would appearto prefer to risk their lives rather than their peer-groupreputations for courage.

Although the propensity to take risks is widelyassumed to vary with circumstances and individuals, thereis no way of testing this assumption by direct measure-ment. There is not even agreement about what units ofmeasurement might be used. Usually the assumption istested indirectly by reference to accident outcomes; onthe basis of their accident records, young men are judgedto be risk seeking, and middle-aged women to be riskaverse. But this test inevitably gets muddled up withtests of assumptions that accidents are caused by errorsin risk perception, which also cannot be measured direct-ly. If Dale Earnhart crashes at 200 mph in his racingcar, it is impossible to determine "objectively" whetherit was because he made a mistake or because he was takinga risk and was unlucky.

Beyond the realm of purely financial risk, both therewards of risk and accident losses defy reduction to acommon denominator; this renders unworkable the economist'spreferred method for performing the risk-management balanc-ing act--cost/benefit analysis. The rewards come in avariety of forms: money, power, glory, love, affection,

self-respect, revenge, curiosity satisfied, or simplythe sensation (pleasurable for some) accompanying a rushof adrenaline. Nor can accident losses be measured witha single metric. Road accidents, the best documented ofall the realms of risk, can result in anything from abent bumper to death, and there is no formula that cananswer the question, How many bent bumpers equal onelife? The search for a numerical measure to attach tothe harm or loss associated with a particular adverseevent encounters the problem that people vary enormouslyin the importance they attach to similar events.Slipping and falling on the ice is a game for childrenand an event with potentially fatal consequences for theelderly.

Figure 9 is a distorted version of Figure 1 withsome of the boxes displaced along an axis labeled"Subjectivity-Objectivity." The box that is displacedfarthest in the direction of objectivity is "balancingbehavior." It is possible to measure behavior directly.It is, for example, well documented that parents havewithdrawn their children from the streets in response totheir perception that the streets have become more dan-gerous.24 It is possible in principle to measure thedecline in the amount of time that children spend in thestreets exposed to traffic, but even here the interpre-tation of the evidence is contentious. Do children nowspend less time on the street because they spend moretime watching television, or do they spend more timewatching television because they are not allowed to playin the streets?25 All of the elements of the risk-com-pensation theory, and those of any contenders of which Iam aware, fall a long way short of the objective end ofthe spectrum. Behavior can be measured, but its causescan only be inferred.

Moreover, risks can be displaced. If motorcyclingwere to be banned in Britain it would save about 500lives a year. Or would it? If it could be assumed thatall the banned motorcyclists would sit at home drinkingtea, one could simply subtract motorcycle accidentfatalities from the total annual road accident deathtoll. But at least some frustrated motorcyclists wouldbuy old clunkers and try to drive them in a way thatpumped as much adrenaline as their motorcycling did, andin a way likely to produce more kinetic energy to bedispersed if they crashed. The alternative risk-takingactivities that they might pursue range from skydivingto glue sniffing, and there is no set of statistics thatcould prove that the country had been made safer, ormore dangerous, by the ban.

Figure 9The Risk Thermostat Stretched

If a road has many accidents it might fairly becalled dangerous; but using past accident rates to esti-mate future risks can be positively misleading. There aremany dangerous roads that have good accident recordsbecause they are seen to be dangerous--children are for-bidden to cross them, old people are afraid to cross them,and fit adults cross them quickly and carefully. The goodaccident record is purchased at the cost of community sev-erance--with the result that people on one side of a busyroad tend no longer to know their neighbors on the other.But the good accident record gets used as a basis for riskmanagement. Officially--"objectively"--roads with goodaccident records are deemed safe and in need of no meas-ures to calm the traffic.

The Dance of the Risk Thermostats

Figure 10 is an attempt to suggest a few of the com-plications confronting those who seek objective measures ofrisk that flow from the reflexive nature of risk. Theworld contains more than 5.5 billion risk thermostats.Some are large--presidents with fingers on buttons--mostare tiny--children chasing balls across streets.Governments and big businesses make decisions that affectmillions if not billions of people. Individuals for themost part adapt as best they can to the consequences ofthose decisions. The damage that they individually caninflict in return, through the ballot box or market, isinsignificant, although in aggregate they can become forcesto reckon with; the slump in the market for beef in re-

Figure 10The Dance of the Risk Thermostats

sponse to fears of BSE has not only caused losses to thebeef industry, but set off a Europe-wide political chainreaction. Overhanging everything are the forces ofnature, floods, earthquakes, hurricanes, plagues, whicheven governments cannot control, although they sometimestry to build defenses against them. And fluttering aboutthe dance floor are the Beijing butterflies beloved ofchaos theorists: they ensure that the best laid plans ofmice and men "gang aft agley." Figure 10 shows but aninfinitesimal fraction of the possible interactions betweenall the world's risk thermostats; there is not the remot-est possibility of ever devising a model or building acomputer that could predict accurately all the consequencesof intervention in this system. The broken line symbol-izes the uncertain impact of human behavior on nature.

The winged creature at the top left was added inresponse to a Time magazine survey (December 27, 1993)that revealed that 69 percent of Americans believe inangels and 46 percent believe they have their own guardianangel. The "angel factor" must influence many risk-takingdecisions--in mysterious ways.

A small part of the dance can be observed directly incrowded local shopping streets on any Saturday morning ascyclists, pedestrians, cars, trucks, and buses all contend

for the same road space. But not all the dangers con-fronting the participants in this dance are visible to thenaked eye. Some cyclists can be observed wearing masks tofilter the air. Lead, oxides of nitrogen, carbon monox-ide, volatile organic compounds, and ultra-fine particu-lates are all invisible substances that some of the betterinformed shoppers might worry about. Few of those worriedpeople will be toxicologists capable of judging the dan-gers directly. Their concerns will usually be based onscientific pronouncements filtered through the media andperhaps augmented by the campaigns of environmentalists.

The dance of the thermostats with its multiple con-nections can also be viewed as a description of the way inwhich an infectious disease is passed among risk takers.Infectious diseases are important to the discussion ofrisks perceived by science because science has an impres-sive record in reducing dangers from infectious diseases.For example, the insight that cholera is spread throughsome water supplies led to closing down sources of contam-inated water and improved sanitation that controlled thespread of the disease more than 30 years before theresponsible bacterium was identified by laboratory scien-tists.26 The spread of sanitation efforts, combined withvaccination, has virtually eradicated cholera from thedeveloped world.

The risk thermostat still plays a part in responsesto infectious disease risks. Now many tourists, protectedby vaccination, venture into parts of the world where theywould previously have feared to go--thereby exposing them-selves to other diseases and dangers. There are numerousother examples of science's defeating risks only for peo-ple to reassert their determination to take risks. TheDavy lamp was heralded as a safety device that reduced thedanger of explosions in mines because it operated at atemperature below the ignition point of methane. But itpermitted the expansion of mining into methane-rich atmos-pheres and was followed by an increase in mining produc-tivity along with explosions and fatalities in themethane-laden environments. Since improvements in braketechnology, when fitted to cars, usually result in driv-ers' going faster, or braking later, the potential safetybenefit gets consumed as a performance benefit.

The Meaning of "Probability"

The probabilities that scientists attach to accidentsand illnesses, and to the outcomes of proposed treatments,

are quantitative, authoritative, confident-sounding expres-sions of uncertainty. They are not the same as the proba-bilities that can be attached to a throw of a pair ofdice. The "odds" cannot be known in the same way, becausethe outcome is not independent of previous throws. Whenrisks become perceptible, when the odds are publicly quot-ed, the information is acted upon in ways that alter theodds.

One form that this action might take is new researchto produce new information. Britain's chief medical offi-cer of health (Sir Kenneth Calman) says, "It is possiblefor new research and knowledge to change the level ofrisk, reducing it or increasing it."27 This view sitsuncomfortably alongside the Royal Society's view28 of riskas something "actual" and capable of "objective measure-ment." If risk is "actual" and subject to "objectivemeasurement," how will further research modify it?

This phenomenon might be described as the Heisenbergproblem.29 The purpose of measuring risk is to provideinformation that can guide behavior. Statements aboutrisk are statements about the future. Accident statis-tics, the most commonly used measure of risk, are state-ments about the past. To the extent that the informationconveyed by accident statistics is acted upon, the futurewill be different from the past. The act of measurementalters that which is being measured.

As scientists, insurance company actuaries, and otherrisk specialists are successful in identifying and publi-cizing risks that have previously been shrouded in igno-rance, they shift them into the directly perceptible cate-gory--and people then act upon this new information. Riskis a continuously reflexive phenomenon; we all, routinely,monitor our environments for signs of safety or danger andmodify our behavior in response to our observations--there-by modifying our environment and provoking further roundsof responses ad infinitum. For example, the more highwayengineers signpost dangers such as potholes and bends inthe road, the more motorists are likely to take care inthe vicinity of the now-perceptible dangers, but also themore likely they are to drive with the expectation thatall significant dangers will be signposted.

What Calman perhaps meant when he said that newresearch might change the level of risk is that the proba-bilities intended to convey the magnitude of the scien-tist's uncertainty are themselves uncertain in ways thatcannot be expressed as probabilities. He should perhaps

have said that a scientific risk estimate is the scien-tist's "best guess at the time but subject to change inways that cannot be predicted." This brings us to uncer-tainty and the cultural construction of risk.

Virtual Risk

We do not respond blankly to uncertainty; we imposemeanings on it. Those meanings are virtual risks.Whenever scientists disagree or confess their ignorance,the lay public is confronted by uncertainty. Virtualrisks may or may not be imaginary, but they have real con-sequences--people act on the meanings that they impose onuncertainty.

The 1995 contraceptive pill scare in Britain is anexample of a "scientific" risk assessment spilling overinto the virtual category. Britain's Committee on theSafety of Medicines issued a public warning on the basisof preliminary, unpublished, non-peer-reviewed evidencethat the new third-generation pill was twice as likely tocause blood clots as the second-generation pill. Theresult was a panic in which large numbers of women stoppedtaking the new pill, with the further result that therewere an estimated 8,000 extra abortions and an unknownnumber of unplanned pregnancies. The highly publicizedtwofold increase in risk amounted to an estimated doublingof fatalities from two to four a year.30 Even when dou-bled, the mortality risk was far below that for abortionsand pregnancies. Such minuscule risks are statisticalspeculations and cannot be measured directly. Subsequentresearch cast doubt on the plausibility of any additionalrisk associated with the new pill. The lesson that thechief medical officer of health drew from this panic(i.e., behavioral response to new information) in hisannual report was that "there is an important distinctionto be made between relative risk and absolute risk." 31

Just whom he held responsible for the failure to appreci-ate this distinction--the public or the government's med-ical experts--he did not make clear.

Perhaps a more important lesson is that scientists,by combining uncertainty with potentially dire conse-quences, can frighten large numbers of people. Dressingup their uncertainties in very low absolute probabilitiesdoes not seem to help--especially when they are presentedin a hastily called press conference that begins with theadvice "don't panic." Calman observed that "although theincreased risk was small, women did need to be informed

that there was a difference in risk between the oral con-traceptives available to them" and that "the message, tocontinue to take the oral contraceptive pill, seemed to beignored in the pressure for action." From where, he mighthave asked himself, did this pressure for action come?Why, women might sensibly ask themselves, are they givingus this new information with such a sense of urgency ifthey expect us to take no action?

The women who stopped taking the pill were imposingmeaning on the uncertainty of the British medical estab-lishment. This uncertainty was projected through andamplified by the media. The fact of the hastily convenedpress conference, the secretive procedures by which theCommittee on the Safety of Medicines and other governmentagencies arrive at their conclusions, and histories ofgovernment cover-ups of dangers such as radiation and madcow disease have resulted in a very low level of publictrust in government. A recent British survey that askedpeople whether they would trust institution X to tell themthe truth about risks found that only 7 percent wouldtrust the government, compared to 80 percent who said theywould trust environmental organizations.32 This mistrustfeeds a paranoid tendency that can hugely exaggerate triv-ial dangers.

Cultural Filters

We all, scientists included, perceive virtual risksthrough different "cultural filters."33 The discovery ofthe Antarctic ozone hole was delayed by a physical equiva-lent of such a filter. U.S. satellites failed to pick upearly indications of the hole because programmers hadinstructed the satellite computers to reject data outsidea specified range as errors. As a result, the low read-ings were trashed as errors.34

The influence of filters can also be detected in thedebate about the effects of low-level radiation. Despitethe accumulation of many decades of evidence, there isstill no agreement about whether there is a safe dose, orperhaps even a therapeutic dose. An article in the April1997 issue of Chemistry in Britain states that

large epidemiological studies for radon levels inparts of the US, Sweden, Finland and China showthat the incidence of lung cancer actuallydecreases with increasing radon exposures, evenfor levels of up to 300 Bq m-3 . . . even inCornwall and Devon, where soils and houses con-

tain the highest levels of uranium and radon inthe UK . . . the number of lung cancers is lowerthan in most other regions of the UK--despitethe fact that the southwest includes a high pro-portion of cigarette smokers.35

This provoked a strong reply from G. M. Kendall andC. R Muirhead of Britain's National Radiological ProtectionBoard, who insisted that radon caused about 2,000 deaths ayear in Britain and suggested that the effect in Devon andCornwall was probably obscured because low smoking rates there caused lower cancer rates even in the presence ofhigh radon levels. But neither side of the argument pre-sented any statistics on smoking in Devon and Cornwall.

John Graham, vice-president in charge of environment,safety and health for British Nuclear Fuels Inc., takesthe argument one step further, advancing the hypothesisthat low-level radiation can have beneficial effects.36 Heargues that background radiation routinely causes cell dam-age, for which effective repair mechanisms exist, and thatthere are optimum exposure levels at which the stimulationof the repair mechanisms outweighs the damage. This layspectator judges the debate to be still unresolved.

When scientists do not know or cannot agree about the"reality" of risks, people are liberated to argue frombelief and conviction. Thompson, Ellis, and Wildavsky37

identified four distinctive "rationalities" of belief andconviction through which people view the world and managetheir risks. Each of the four rationalities is associatedwith a "myth of nature," illustrated by the behavior of aball in a landscape (Figure 11).

As described in Table 2, individualists see nature asbenign (and robust); it can be made to do what humans com-mand. Egalitarians see nature as ephemeral; nature is tobe obeyed. Hierachists see nature as alternately perverseand tolerant; with proper attention to rules, nature canbe managed. Fatalists see nature as capricious; naturehas the upper hand; there's nothing to be done.

Where do average people fall on Figure 11 and Table2? Depending on context and circumstance, they can findthemselves in any or all of the categories.

Table 2Rationalities of Belief and Their Associated Myths of Nature________________________________________________________________________________________

Rationality Myth of Nature________________________________________________________________________________________

_________________________________________________________________________

• Individualists: nature benignIndividualists are enterprising "self-made" people,relatively free from control by others, who strive toexert control over their environment and the peoplein it. Their success is often measured by wealthand the number of followers they command. Theself-made Victorian mill owner and present-dayventure capitalist are good representatives of thiscategory. They oppose regulation and favor freemarkets. Nature is to be commanded for humanbenefit.

• Egalitarians: nature ephemeralEgalitarians have strong group loyalties but littlerespect for externally imposed rules, other thannature's. Their central rule is the precautionaryprinciple, considered necessary to protect naturefrom human abuses. Group decisions are arrived atdemocratically and leaders rule by force of person-ality and persuasion. Members of religious sects,communards, and environmental pressure groups allbelong to this category. Nature is to be obeyed.

• Hierarchists: nature perverse/tolerantHierarchists inhabit a world with strong groupboundaries and binding prescriptions. Social rela-tionships in this world are hierarchical and every-one knows his or her place. Members of caste-bound Hindu society, soldiers of all ranks, and civilservants are exemplars of this category. Nature isto be managed.

• Fatalists: nature capriciousFatalists have minimal control over their own lives.They belong to no groups responsible for the deci-sions that rule their lives. They are nonunionizedemployees, outcasts, refugees, untouchables. Theyare resigned to their fate and see no point inattempting to change it. Nature, they expect, willthrow things at them, and the best they can do isduck if they see something coming.

Nature benign is represented by a ball in a cup(see Figure 11): nature, according to this myth, ispredictable, bountiful, robust, stable, and forgivingof any insults humankind might inflict upon it;however violently it might be shaken, the ballcomes safely to rest in the bottom of the basin.Nature is the benign context of human activity, notsomething that needs to be managed. The individ-ualist's management style—relaxed, exploitative,laissez-faire—fits this myth.

Nature ephemeral is represented by a ball bal-anced precariously on an overturned cup: herenature is fragile, precarious, and unforgiving. It isin danger of being provoked by human greed orcarelessness into catastrophic collapse. People,the myth insists, must tread lightly on the earth.The egalitarian's guiding management rule is theprecautionary principle; it is necessary to obeynature.

Nature perverse/tolerant combines modified ver-sions of the first two myths. Within limits, naturecan be relied upon to behave predictably. It is for-giving of modest shocks and can look after itselfin minor matters. Care must be taken not to knockthe ball over the rim; regulation is required to pre-vent major excesses. This is the ecologist's equiv-alent of a mixed-economy model. The hierarchistmanager's style is interventionist.

Nature capricious: nature is unpredictable. Theappropriate management strategy is laissez-faire,in the sense that there is no point to management.Where adherents to the myth of nature benigntrust nature to be kind and generous, the believerin nature capricious is agnostic; the future mayturn out well or badly, but in any event, it isbeyond his control. The fatalist's nonmanagementmotto is que sera sera.

Figure 11Four Rationalities: A Typology of Bias

Source: Adams, Risk.

Coping with Risk and Uncertainty: The Dose-Response Curve

Wherever the evidence in a dispute is inconclusive,the scientific vacuum is filled by the assertion of con-tradictory certitudes. There are numerous risk debates,such as those about the relationship between BSE and CJD,in which scientific certainty is likely to be a rare com-modity for the foreseeable future. Issues of health,safety, and the environment--matters of life and death--will continue to be decided in the absence of conclusivescientific knowledge.

Just how remote is the prospect of scientific resolu-tion, and how large is the scientific vacuum, can beillustrated with the help of some numbers taken from areport by the U.S. National Research Council,38 which notesthat about 5 million different chemical substances areknown to exist and that the risks from every one are theo-retically under federal government regulatory jurisdiction.In 1983, when the report was first published, fewer than30 chemicals had been definitely linked to cancer in

humans, and about 7,000 had been tested for carcinogenici-ty in animals.

Even allowing for the advances of cancer researchsince 1983, these last two numbers almost certainly great-ly exaggerate the extent of existing knowledge. Given theethical objections to direct testing on humans, tests forcarcinogenicity are done on animals. The NRC reportobserves, "There are no doubt occasions in that observa-tions in animals may be of highly uncertain relevance tohumans"; it also notes that the transfer of the results ofthese tests to humans requires the use of scaling factorsthat can vary by a factor of 35 depending on the methodused and observes that "although some methods for conver-sion are used more frequently than others, a scientificbasis for choosing one over the other is not estab-lished."39 A further difficulty with such experiments isthat they use high doses in order to produce results thatare clear and statistically significant for the animalpopulations tested. But for every toxic chemical, thedose levels at issue in environmental controversies areprobably much lower.

A mathematical model is necessary to extrapolate fromthe effects at high dose levels at which effects are unam-biguous for animals to the much lower exposures experi-enced by the general human population. Figure 12 illus-trates the enormous variety of conclusions that might bedrawn from the same experimental data, depending on theassumptions used in extrapolating to lower doses. In manycases, there is only a single data point--almost nevermore than two--from an animal test. Those points, whichare related to numbers of tumors at high doses, anchor allthe extrapolation lines so that, as shown on Figure 12,the estimates produced by the five different models are inreasonable agreement in the upper right-hand corner of thegraph. The models agree that high dose levels producehigh response levels. But the models diverge at lowerdoses, and at doses that are typical of human exposurelevels, the estimates of risk can differ by factors of10,000 or more. Thus, cultural filters, not science, dic-tate choices among the models.

• The supralinear model assumes that the level ofrisk will remain high as dose levels are reduced.

• The linear model, preferred by U.S. regulatoryagencies for estimating the risks from all carcinogensand by regulatory agencies in other countries forestimating risks from carcinogens that directly affect

Figure 12Alternative Dose-Response Extrapolations from the Same Empirical Evidence

Source: National Research Council, Risk Assessment in the Federal Government, p. 263.

DNA,40 assumes that there is a direct relationshipbetween dose and risk. Reducing the dose by a factorof two reduces risk by the same amount.

• The two sublinear models assume that reducing thedose by a factor of two reduces the risk by a greaterfactor. Such models are often proposed by regulatedindustries in the United States, but they have notbeen adopted by regulatory agencies.

• The threshold model assumes that risk falls to zerowhen the dose levels fall below a certain value, thethreshold dose. Threshold models are used by regula-tory agencies in European countries to estimate risks

for carcinogens that do not directly affect DNA, butthey are rarely used in the United States.41

Four other sources of uncertainty are of even greatersignificance in making risk estimates. First, variabilityin susceptibility within exposed human populations, com-bined with the variability in their levels of exposure,makes predictions of the health effects of substances atlow dose levels a matter of guesswork. Second, the longlatency period between exposures to most carcinogens--suchas cigarettes and radiation--and the occurrence of cancermakes the identification of many carcinogens and theircontrol impossible before the exposure of the public.Third, the synergistic effects of substances acting incombination can make innocent substances dangerous, and themagnitude of the number of combinations that can be creat-ed from 5 million substances defies the capabilities ofall known computers. And fourth, the gremlins exposed bychaos theory (represented in Figure 10 by the Beijing but-terfly) will always confound the seekers of certainty incomplex systems sensitive to initial conditions.

Figure 13 shows the risk thermostat fitted with cul-tural filters.42 The mythological figures of cultural the-ory are caricatures, but they have numerous real-lifeapproximations in debates about risk. Long-running contro-versies about large-scale risks are long running becausethey are scientifically unresolved and unresolvable withinthe time scale imposed by the necessity of making deci-sions. This information void is filled by people who rushin from the four corners of cultural theory's typologyasserting their contradictory certitudes. The clamorousdebate is characterized not by irrationality but by pluralrationalities.

The contending rationalities not only perceive riskand reward differently, but they also differ according tohow the balancing act ought to be performed.

• Individualists scorn authority as the "nanny state"and argue that decisions about whether to wear seatbelts or eat beef should be left to individuals andsettled in the market.

• Egalitarians focus on the importance of trust; riskmanagement is a consensual activity; consensus build-ing requires openness and transparency.

Figure 13The Risk Thermostat Fitted with Cultural Filters

• Hierarchists are committed to the idea that themanagement of risk is the job of authority--approp-riately assisted by expert advisers. They cloaktheir deliberations in secrecy or technical mumbo-jumbo43 because the ignorant lay public cannot berelied upon to interpret the evidence correctly oruse it responsibly.

• Fatalists take whatever comes along.

Mad Cow Disease

The mad cow disease (or more formally, the bovinespongiform encephalopathy/Creutzfeldt-Jakob disease,BSE/CJD) controversy contains all of the sources of uncer-tainty discussed with respect to Figure 12. There hasbeen uncertainty and great controversy about whether the"new strain" of CJD that affects young people (called"vCJD" or "variant CJD" to distinguish it from "spCJD"that occurs sporadically in older people) is a human formof BSE. The very existence and nature of prions, the

hypothesized agents of transmission of the disease, washotly disputed,44 and arguments raged about whether prionsare composed only of proteins.45 Now, however, a scientif-ic consensus is emerging about the causative role of pri-ons in diseases such as BSE and CJD.46

Up until now BSE/vCJD research has concentrated ondemonstrating the possibility of BSE jumping the speciesdivide from cows to humans. Estimation of the dose levelsat which it becomes a significant threat and identifica-tion of the causes of variation in susceptibility arestill projects to be worked on. The long and variablelatency period for spongiform diseases makes reconstructionof past exposures to presumed causes and exploration ofhypotheses about synergistic effects extremely difficult.Great uncertainty surrounds the origins of vCJD, and vCJDremains an extremely rare disease. With all that uncer-tainty, it's no surprise that people respond differentlyto the limited information that is available.

Individualists, assiduous collectors of information,are comfortable with uncertainty. Their optimism makesthem gamblers--they expect to win more than they lose.Markets in their view are institutions with a record ofcoping with uncertainty successfully. If the experts can-not agree about BSE, there is no basis upon which centralauthority can act; the risk should be spread by lettingindividual shoppers decide for themselves.

The egalitarian instinct in the face of uncertaintyis to assume that authority is covering up somethingdreadful and that untrammeled markets will create somethingworse. Egalitarians favor democratizing the balancing actby opening up the expert committees to lay participationand holding public inquiries to get at the truth, whichthey expect will show nature to be precariously balancedon the brink of disaster. In such cases, the precaution-ary principle must be imposed to protect nature, and theprecautionary principle, which calls for no change unlessthere is no possibility of adverse outcomes from it, jus-tifies the draconian intervention in markets that theyfavor. While hostile to external regulation, they arehappy to impose strict rules of behavior upon themselvesand others to fend off catastrophe.

Ignorance is a challenge to the very idea of authori-ty and expertise. The response of hierarchists is to con-ceal their doubts and present a confident public face.Confession of ignorance or uncertainty does not come easi-ly to authority. In the face of uncertainty about anissue such as BSE, they seek to reassure.47 The spectacle

of the minister of agriculture, fisheries, and food'sforcing his daughter to eat a hamburger as a public demon-stration of his belief in the safety of British beef wasfollowed by the government-forced slaughter of 2 millioncows. It is difficult to imagine that the minister wasless certain of the safety of his action than the otherdecisionmakers were about the risk when they ordered theslaughter.

The fatalists just carry on, drinking beer, readingUSA Today, and buying lottery tickets. They might acceptinvitations to buy a rib roast if an individualist offer-ed.

The BSE/vCJD affair demonstrates the provisionalnature of lay reactions to official information. When thestory first broke, sales of beef plummeted. Sales felleven more in Germany, where the German government main-tained that there was no BSE, than in Britain. A year anda half later, when the British government banned the saleof beef on the bone as an additional precaution, the news-papers were full of stories of people rushing to purchasethe proscribed meat before the ban came into effect.48

This perhaps demonstrates the difficulty that the newsmedia have in maintaining a state of alarm in the absenceof a high body count.

As recently as the summer of 1997 the British MedicalJournal summed up the state of knowledge thus: "We do notknow how or indeed if bovine spongiform encephalopathy istransmitted to humans." 49 One of the report's "key mes-sages" was that "the observation of a group of compara-tively young patients with Creutzfeldt-Jakob disease char-acterized by unusual neuropathological features during1994-96 remains unexplained."

At that same time, a leading researcher in the field,Professor John Collinge, proclaimed in an interview withthe Times (of London) medical correspondent (August 7,1997) that "CJD could become an epidemic of biblical pro-portions" (this dramatic quotation served as the headlinefor the article). Professor Collinge went on to say,

I am now coming round to the view that doctorsworking in this field have to say what theythink, even though this may give rise to anxi-eties which later turn out to be groundless.. . . We have to face the possibility of a dis-aster with tens of thousands of cases . . . wejust don't know if this will happen, but what is

certain is that we cannot afford to wait andsee.50

Three days later, the Sunday Telegraph published arobust individualist response by Christopher Booker toProfessor Collinge's egalitarian call for precautionaryaction in the face of uncertainty. The response alsoraised the question of what the nation could afford:

The efforts of the scientists behind last year'sBSE scare to defend their alleged link with "newvariant Creutzfeldt Jacob disease" become evermore comical as the epidemic they promised failsto materialize. . . . How much longer should wecontinue to look for objective guidance on thismatter to experts who have invested so much oftheir own personal reputations in the theory thata link between BSE and new variant CJD exists?Faced with a bill now rising above £5 billion. . . how much longer can we afford it?51

There has been no epidemic of biblical proportions;there was a total 23 cases of vCJD as of January 1, 1998,in the UK, and no patterns in terms of occupation or eat-ing habits have been found among the 23.52 The absence ofcommon habits or exposures among the afflicted peopleweakens any connection that can be drawn between mad cowdisease and CJD. But the numbers are very small, theanalysis is inconclusive, and there is a possible incuba-tion period of unknown length.

Who knows about the future? A news article in Natureneatly captured the "on the one hand, on the other hand"judgment of experts: "The rate of new cases is notincreasing . . . but it may be several years before we canbe confident that this is not a period of comparative calmbefore a storm."53

Toward the end of 1997 a scientific consensusappeared to be emerging in support of BSE's being thecause of vCJD. But more recently Stanley Pruisner, thewinner of the Nobel Prize for his work on prions, pushedthe issue firmly back into the realm of virtual risk. Thefollowing are excerpts from evidence he presented to theofficial British government inquiry into BSE on June 6,1998. (The transcript of his evidence is available atwww.bse.org.uk.)

Commenting on the evidence for BSE's causing vCJD hesaid, "I simply do not understand what all this means. I

Table 3BSE/CJD: A Typology of Bias_________________________________________________________________________________________________________

Fatalist Hierarchist

_______________________________________________________________________________________

•· "They should shoot the scientists, not cull the calves.Nobody seems to know what is going on." Dairy farmerquoted in the Times, August 2, 1996.

• "Charles won't pay for Diana's briefs." Main headlinein the Sun on March 21, 1996, the day every other paperled with the BSE story.

• "We require public policy to be in the hands of electedpoliticians. Passing responsibility to scientists can onlyundermine confidence in politics and science." John Durant,The Times Higher Education Supplement, April 5, 1996.• "As much as possible, scientific advice to consumersshould be delivered by scientists, not politicians." TheEconomist, March 21, 1996.• "I believe that British beef is safe. I think it is good foryou." Agriculture Minister Douglas Hogg, December 6,1995.• "I believe that lamb throughout Europe is wholly safe."Douglas Hogg, July 23, 1996.• "I felt the need to reassure parents." Derbyshire educationchief quoted in the Sun, March 21, 1996.• "I have not got a scientific opinion worth listening to. Myjob is simply to make certain that the evidence is drawn tothe attention of the public and the Government does what weare told is necessary." Health Secretary Stephen Dorrel,Daily Telegraph, March 22, 1996.• "We felt it was a no-goer. MAFF Ministry of Agriculture,Fisheries, and Food, U.K. already thought our proposals werepretty 'radical.'" Richard Southwood, explaining why he hadnot recommended a ban on cattle offal in human food in1988. Quoted by B. Wynne, Times Higher EducationSupplement, April 4, 1996.

Individualist• "The precautionary principle is favored by environmentalextremists and health fanatics. They feed off the lack of sci-entific evidence and use it to promote fear of the unknown."T. Corcoran, Toronto Globe and Mail, March 27, 1996.• "I want to know, from those more knowledgeable than I,where a steak stands alongside an oyster, a North Sea mack-erel, a boiled egg and running for the bus. Is it a chance in amillion of catching CJD or a chance in ten million? I amgrown up. I can take it on the chin." Simon Jenkins, theTimes, quoted by J. Durant in Times Higher EducationSupplement, April 5, 1996.• "'Possible' should not be changed to 'probable' as has hap-pened in the past." S.H.U. Bowies, FRS, the Times, August12, 1996.• "It is clear to all of us who believe in the invisible hand ofthe market place that interference by the calamity-promotingpushers of the precautionary principle is not only hurtful butunnecessary. Cost-conscious non-governmental institutionsare to be trusted with the protection of the public interest." P.Sandor, Toronto Globe and Mail, March 27, 1996.• "I shall continue to eat beef. Yum, yum." Boris Johnson,Weekly Telegraph, no. 245.

Egalitarian• Feeding dead sheep to cattle, or dead cattle to sheep, is "unnat-ural" and "perverted." "The present methods of the agriculturalindustry are fundamentally unsustainable." "Risk is not actuallyabout probabilities at all. It's all about the trustworthiness of theinstitutions which are telling us what the risk is." MichaelJacobs, The Guardian, July 7, 1996.• "The Government . . . choose to take advice from a smallgroup of hand-picked experts, particularly from those who thinkthere is no problem." Lucy Hodges, Times Higher EducationSupplement, April 5, 1996.• "It is the full story of the beginnings of an apocalyptic phe-nomenon: a deadly disease that has already devastated thenational cattle herd . . . could in time prove to be the most insidi-ous and lethal contagion since the Black Death." The "BritishGovernment has at all stages concealed facts and corrupted evi-dence on mad cow disease."• "Great epidemics are warning signs, symptoms of disease insociety itself." G. Cannon in the foreword to Mad Cow Diseaseby Richard Lacey.• "My view is that if, and I stress if, it turns out that BSE can betransmitted to man and cause a CJD-like illness, then it would befar better to have been wise and taken precautions than to havenot." Richard Lacey, ibid.

do not know that this tells us that variant CJD comes fromBSE."

What advice could he offer on the safety of beef?

I have worked in this field for 25 years. Andbefore there was ever BSE I mainly worked onscrapie [a spongiform encephalopathy of sheep],because we did very little on human CJD in theinitial phases of the work. Did I go out andeat lamb chops; did I go out and eat lamb brainand sheep brain? The answer was "no," but itwas based not on scientific criteria. It wasbased on just emotion. It is what I said earli-er. When there is a disease like BSE things donot sound appetizing. But at a scientificlevel, I cannot give you a scientific basis forchoosing or not choosing beef, because we do notknow the answers.54

Table 3, arranged in the same fashion as Figure 11,presents a representative selection of comments about BSEcategorized by cultural theory typology.

• The individualist will continue to believe hisrisks are small and manageable and want to be leftalone with the information that he has or wants.

• The egalitarian will press for more intensive sur-veillance, more oversight and public participation,and, perhaps, more regulations to guard against thepotential "storm."

• The hierarchist may suffer the most; the fallingout between political and scientific authority mani-fest in the upper right-hand corner of Figure 1 ischaracteristic of the disarray into which hierarchyfalls when its mask of authoritative knowledge istorn off--the ball in the top right-hand corner hasgone over the rim. He will reassure the public,while commissioning further research as a precaution-ary measure.

• The fatalist will be the least concerned, perhaps,because he washed his hands of the whole affair along time ago, if he noticed it at all. Or becausehe reads the Sun.

Should We Follow a Risk-Averse Environmental Policy?

Who are "we"? "Risk-averse" and "risk-seeking" areusually labels that people apply to others whose riskthermostats are fitted with different cultural filters.Those who argue for a more risk-averse policy are, ineffect, saying that there is a discrepancy between thedangers that they perceive and the risks that they areprepared to take. The activities of environmental groups(egalitarians) lobbying for the precautionary principle canbe seen as a collective behavioral response to this dis-crepancy.

The environmentalist case rests on the conviction thatgrowth processes--economic and demographic--are pressingagainst global limits. Perhaps the best exemplars of thisconviction are the Club of Rome authors who argue inBeyond the Limits that

the human world is beyond its limits. Thefuture, to be viable at all, must be one ofdrawing back, easing down, healing. . . . Themore we compiled the numbers, the more they gaveus that message, loud and clear.55

In the BSE debate the complementary message that isreceived and retransmitted loud and clear by egalitariansis that BSE is a punishment for unnatural methods of agri-culture. Modern intensive, high-energy production methods,veal crates, battery chickens, genetic manipulation, foodpreservation methods, pesticides, and feeding meat to her-bivores are all, according to this perspective, aspects ofthe same hubristic syndrome. The remedy? Nature is to beobeyed; we must (re)turn to more humane and extensive,organic, natural methods of production.

This message is countered by an individualist backlashthat views the environmental lobby itself as an environ-mental threat. Julian Simon, for example, insists thatthere is a positive correlation between indices of materi-al wealth and an improving environment. With Herman Kahn,he has argued,

We are confident that the nature of the physicalworld permits continued improvement in humankind's economic lot . . . indefinitely.. . . There are always newly arising local prob-lems, shortages, and pollutions. . . . But thenature of the world's physical conditions and theresilience in a well-functioning economic andsocial system enable us to overcome such prob-

lems, and the solutions usually leave us betteroff than if the problem had never arisen; thatis the great lesson to be learned from humanhistory.56

This "rationality," when confronted with the evidenceof BSE/CJD, sees no evidence of serious harm. It pointsto the enormous benefits of intensive agricultural produc-tion: the freedom from toil and drudgery provided by mod-ern machinery, the improved nutrition and material stan-dards of living enjoyed by both farmers and consumers, thevast range of choice now available to food shoppers.Their version of the precautionary principle sees allthese benefits being placed in jeopardy by an overreactionto tenuous scientific evidence about the cause of a veryrare illness.

One side, embracing the precautionary principle, saysthat if you cannot prove it is safe, you must treat it asdangerous. The other side, citing examples such as thefact that aspirin would never have gotten onto the marketif all its real and potential side effects had been known,says that such an approach would quickly bankrupt anyendeavor and argues that if you cannot prove it is danger-ous, you should treat it as safe.

Governments, the hierarchists, are caught in the mid-dle. Committed to the idea that problems such as BSE canbe managed and embarrassed by their manifest failure to doso convincingly, they sought to reassure the public thateating British beef was probably safe and commissionedmore research that they hoped would confirm it. When itdidn't, they initiated a program of mass slaughter, whichthey justified, not on the grounds that it was necessaryto contain the disease, but on the grounds that it wasnecessary to restore public confidence.57

Finally the fatalist, unless he was one of the 23 tofall victim to vCJD or knows someone who did, might knownothing about BSE-vCJD. Noticing that some cuts of beefare missing from the butcher shop, he might ask and findout that nothing much has happened. Perhaps buttressed inhis belief that the whole risk debate is baloney, he mightbuy a lamb chop.

Conclusion

Science has been very effective in reducing uncertain-ty but much less effective in managing it. The scientificrisk literature has little to say about virtual risks--and

where the scientist has insufficient information even toquote odds, the optimizing models of the economist are oflittle use. A scientist's "don't know" is the verbalequivalent of a Rorschach inkblot: some will hear a cheer-ful reassuring message; others will listen to the samewords and hear the threat of catastrophe.

Science has a very useful role in illuminating dan-gers that were previously invisible, and thereby shiftingtheir management into the directly perceptible category.Where science has been successful, it has reduced uncer-tainty, and thereby shrunk the domain of risk perceivedthrough science; now that its causes are well understood,cholera, for example, is rarely discussed in terms ofrisk. But where the evidence is simply inconclusive andscientists cannot agree about its significance, we all,scientists included, are in the realm of virtual risk, inthe realm of hypothesis.

Figure 14 indicates the relative significance that Isuggest hypotheses should be accorded in risk debates.The future is uncertain. What we do not know about itgreatly exceeds what it is ever likely to tell us.

The role of science in debates about risk is firmlyestablished; clearly we need more information and under-standing, of the sort that only science can provide, aboutthe probable consequences of "balancing behaviors" for both"rewards" and "accidents." But equally clearly we mustdevise ways of proceeding in the absence of scientificcertainty about such consequences--science will never haveall the answers--and in so doing we must acknowledge the

Figure 14Reality?


Perceiveddirectly

Virtualrisk

(hypothesis)

scientific elusiveness of risk. The clouds do not respondto what the weather forecasters say about them. People dorespond to information about risks, and thereby changethem.

In the presence of virtual risk, the precautionaryprinciple becomes an unreliable guide to action. Considerthe ultimate virtual risk, discussed from time to time ontelevision and in our newspapers. Edward Teller invokedthe precautionary principle to argue for the commitment ofvast resources to the development of more powerful H-bombsand delivery systems to enable the world to fend offasteroids--even if the odds of their ever being needed areonly one in a million.58 But we are also told by Russia'sdefense minister that "Russia might soon reach the thresh-old beyond which its rockets and nuclear systems cannot becontrolled."59 Which poses the greater danger to life onearth--asteroids or H-bombs and delivery systems out ofcontrol?

Simon, after a robust display of optimism, observesthat nothing has reduced the "doomsayers' credibility withthe press, or their command over the funding resources ofthe federal government."60 Health and environment debateshave a durable and predictable character. The specificissues may change, but the same caricatures from the cul-tural theory typology reappear in each new debate. TheBSE/CJD controversy is but the most recent installment ina much larger, long-running debate. On all sides convic-tions appear to be as strongly held as ever, and asresistant as ever to contrary evidence.

Scientists have cultural filters about the risks theyunderstand as well as the risks they are trying to under-stand. For scientists and lay people alike, our culturalfilters are parts of our identities and essential to oursense of social solidarity. The persistence of contradic-tory rationalities built upon partial knowledge suggeststhat we are doomed, for the foreseeable future, to contin-ue to argue from different premises.

Debates about BSE/CJD, radiation, and asteroid defens-es are debates about the future, which does not existexcept in our imaginations. They are debates to whichscientists have much to contribute, but not ones that canbe left to scientists alone. An understanding of the dif-ferent ways in which people tend to respond to uncertaintycannot settle arguments; but the arguments are likely tobe more civilized, and our cultural filters less crudelyselective, to the extent that we are sensitive to thesedifferences and understand their causes and effects.

In brief, it is important to be clear about thenature of the risk under discussion. We live in an uncer-tain world, but certain conclusions about the managementof risk might, nevertheless, still be ventured.

Where risks are directly perceptible,

• everyone takes risks; everyone is a risk manager;

• taking risks leads, by definition, to accidents;the pursuit of a world free of accidents is a futileexercise;

• it is important to distinguish self-risk (e.g.,driving without a seat belt) from behavior that putsothers at risk (e.g., driving at 100 mph down a busyshopping street); the second is a legitimate area forregulation; the first is not;

• attempts to criminalize self-risk are likely to beworse than useless; they are likely to redistributethe burden of risk in ways that harm innocent thirdparties;

• everyone has a risk thermostat, and he may adjustit so that he has the risk level he likes regardlessof the experts' best efforts to decrease risk;

• institutional risk managers who do not take accountof the reasons that people have for taking and bal-ancing risks--the rewards of risk--will be frustrated.

Where risks are perceived with the help of science,

• science can reduce uncertainty by illuminating theconnection between behavior and consequence;

• science, effectively communicated, can defeat super-stition and purely imaginary scares; but

• science cannot provide "objective" measures of risk;

• risks come in many incommensurable forms that defyreduction to a common denominator;

• the act of measurement alters that which is beingmeasured;

• risk is a reflexive phenomenon; in managing riskswe are continually modifying them; in the realm ofrisk a Heisenberg principle probably rules.

Where scientists don't know or cannot agree,

• we are in the realm of virtual risk where pluralrationalities contend;

• virtual risks are cultural constructs;

• they may or may not be real--science cannot settlethe issue--but they have real consequences;

• the precautionary principle is of no help; differ-ent rationalities adhere to very different versions ofthe principle;

• virtual risks are a fact of life; science willnever have all the answers;

• humility in the face of ignorance is a preconditionfor civilized debate about virtual risks.

Notes

1. As discussed below, however, regulators and otherinstitutional managers of risk often proclaim zero risk tobe their goal.

2. A term coined by Gerald Wilde in "The RiskCompensation Theory of Accident Causation and Its PracticalConsequences for Accident Prevention," Paper presented in1976 to the annual meeting of the ÖsterreichischeGesellschaft für Unfallchirurgie, Salzburg. His mostrecent book on this theme is Target Risk (Toronto: PDEPublications, 1994).

3. Calls by Cato staff to the National Highway TrafficSafety Administration to obtain research results about howmany lives have been saved through seat belt use wereunsuccessful.

4. Leonard Evans, Traffic Safety and the Driver (NewYork: Van Nostrand Rheinhold, 1991), p. 278. The evidenceconcerning the "when-used effectiveness of belts" is basedon crash testing using dummies and on paired-comparisonstudies, which examine the injuries suffered in crasheswhen one occupant is belted and another unbelted."Selective recruitment effects" must be allowed for becausethe timid and cautious are most likely to belt up volun-tarily, while the wild and reckless are most likely todefy a law.

5. Ibid., p. 327.

6. Department of Transport, Press release, October 15,1985.

7. See John Adams, Risk (London: Taylor and Francis,1995), chapter 7.

8. J. Broughton and D. C. Stark, The Effect of the 1983Changes to the Law Relating to Drink/Driving (Crowthorne,U.K.: Transport and Road Research Laboratory, 1986).

9. See Adams, Risk, chapters 4 and 8, for a discussion ofthis phenomenon.

10. See Ortwin Renn et al., "The Rational Action Paradigmin Risk Theories: Analysis and Critique," in Risk in theModern Age: Science, Trust, and Society, ed. Maurice J.Cohen (London: Macmillan, 1998, in press).

11. Britain's National Health Service has produced some-thing similar in Risk Management in the NHS, Department ofHealth, July 1996.

12. Richard Charlton, director of exploration and produc-tion, Shell Oil, "Where the Buck Stops," Shell World,February 1991, p. 8.

13. Koos Visser, head of health, safety, and environment,Shell Oil, "Prudence and the Gambler," Shell World,February 1991, pp. 24-25.

14. Minister Ian Taylor in DTI, Press notice P96/686,September 11, 1996.

15. Editorial in RSS News 24, no. 4 (December 1996): 1.

16. On the State of the Public Health: The Annual Reportof the Chief Medical Officer of the Department of Healthfor the Year 1995 (London: Her Majesty's StationeryOffice, 1996).

17. Road Accident Statistics: Great Britain 1995 (London:Her Majesty's Stationery Office, 1996).

18. These factors are taken from U.S. statistics in Evans.

19. Defined by Evans as a crash "of sufficient severity tokill 80-year-old male drivers" (p. 34).

20. Evans, pp. 36, 91, 146; and John Adams, Risk andFreedom: The Record of Road Safety Legislation (London:Transport Publishing Projects, 1985), p. 6.

21. John Durant, "Overcoming the Fear of Flying with Joe-Public as Co-Pilot," Times Higher Education Supplement,March 14, 1997. "Us" in the context refers, I presume, tohis scientific audience at the Royal Society, not the laypublic.

22. Evans, p. 362, contains a summary of the argument setout in L. Evans, M. C. Frick, and R. C. Schwing, "Is ItSafer to Fly or Drive?--A Problem in Risk Communication,"Risk Analysis 10 (1990): 259-68.

23. F. Weinberg, Letter to Times (London), December 28,1996.

24. In Britain in 1971, 80 percent of 7- and 8-year-oldchildren got to school unaccompanied by an adult; by 1990this number had fallen to 9 percent, with parents givingfear of traffic as the principal reason for curbing theirchildren's freedom. M. Hillman, J. Adams, and J.Whitelegg, One False Move: A Study of Children'sIndependent Mobility (London: Policy Studies Institute,1990).

25. In Britain in the late 1980s, the Department ofTransport distributed a leaflet ("A Lesson for Life:Teaching Road Safety for Parents of 1-15 Year Olds") inprimary schools, advising parents that it would irresponsi-ble to allow any child under the age of 12 out of thehouse unaccompanied by an adult.

26. Dr. John Snow mapped the new cases of cholera in anoutbreak in London in 1849. The focus of the cluster wasthe Broad Street well in Soho. The pump handle wasremoved and the outbreak subsided.

27. On the State of the Public Health, p. 8.

28. Royal Society, Risk: Analysis, Perception andManagement (London: Royal Society, 1992), p. 1.

29. According to Heisenberg's uncertainty principle, theattempt to measure the location of a particle alters itsposition in an unpredictable way.

30. Quoted on "Anxiety Attack," BBC2, June 11, 1997.

31. On the State of the Public Health, p. 9.

32. C. Marris, I. Langford, and T. O'Riordan, "IntegratingSociological and Psychological Approaches to PublicPerceptions of Environmental Risks: Detailed Results from aQuestionnaire Survey," CSERGE Working Paper GEC 96-07,University of East Anglia, 1996.

33. See Adams, Risk, chapter 3, "Patterns in Uncertainty."

34. R. E. Benedick, Ozone Diplomacy (Cambridge, Mass.:Harvard University Press, 1991), p. 19.

35. Eric Hamilton, "Radon," Chemistry in Britain, April1997, p. 49. (300 Bq m-3 is equivalent to about twice theEnvironmental Protection Agency's "level of concern" forradon in homes.)

36. See Bernard L. Cohen, "Lung Cancer Rate vs. Mean RadonLevel in U.S. Counties of Various Characteristics," HealthPhysics 72 (1997): 114-19, for an example of an analysisthat supports the idea that certain, low levels of radonexposure have beneficial health effects.

37. Michael Thompson, R. Ellis, and A. Wildavsky, CulturalTheory (Boulder, Colo.: Westview, 1990).

38. National Research Council, Risk Assessment in theFederal Government: Managing the Process (1983; Washington:National Academy Press, 1992).

39. Ibid., p. 27. Since the NRC report was published, theEPA and the FDA, which most depend on animal tests to jus-tify their regulations, have reconciled their differencesabout scaling factors. They took an average of their com-peting values. See Environmental Protection Agency, "ACross-Species Scaling Factor for Carcinogen RiskAssessments Based on Equivalence of mg/kg3/4/day," FederalRegister 57 (1992): 24152-73.

40. See Michael Gough, "Science Policy Choices andEstimation of Cancer Risk Associated with TCDD," RiskAnalysis 8 (1988): 337-42.

41. Michael Gough and Stephen Milloy, "EPA's Cancer RiskGuidelines: Guidance to Nowhere," Cato Policy Analysis no.263, November 12, 1996, p. 25.

42. The method of presenting information in Figure 12might be considered both as the product of a cultural fil-ter and as a cultural filter in its own right. On such a

graph it is not possible to show beneficial effects, onlyharmful effects approaching zero. Why, one wonders, whenvirtually all of the therapies produced by the pharmaceu-tical industry, including aspirin, are toxic above certaindoses and beneficial below certain doses, should the con-ventional dose-response curve preclude the possibility of abenign effect? The answer, perhaps, lies in the divisionof labor discussed above in the section on institutionalmanagement of risk. The responsibility of most risk man-agers is to focus on the bottom loop of Figure 1, to tryto minimize the number and magnitude of adverse outcomes.Thus the first question that the U.S. Food and DrugAdministration or the British Committee on the Safety ofMedicines will ask of a new food or drug is whether it hasharmful effects. The emphasis of the manufacturers, thefood and drug companies, is likely to be on the top loop,the rewards to the customer and the profits to themselves.For medical risks there is a dearth of risk managementinstitutions that seek to strike a balance between poten-tial adverse and beneficial consequences.

43. Wendy E. Wagner, "The Science Charade in Toxic RiskRegulation," Columbia Law Review 95 (1995): 1613-1723.

44. Robert Rohwer, quoted in "Nobody Has Proven That ThesePrions Really Exist," Special News Report, Science,December 7, 1996. "The prion hypothesis is the 'coldfusion' of infectious disease--it's a very radical idea,and just like cold fusion it has some very appealingaspects. But because it's so radical it deserves a veryhigh level of scepticism and scrutiny before it's adopt-ed."

45. "Not the Last Word on the BSE Crisis," editorial,Nature 389 (October 2, 1997): 423

46. Jeffrey Almond and John Pattison, "'Protein Only'Prions," Nature 389 (October 2, 1997): 438. For a popularaccount of the controversy, see Richard Rhodes,"Pathological Science," New Yorker, December 1, 1997,pp. 34-49.

47. The propensity of authority to cope with ignorance bydenying its existence is described by Jerome Ravetz in"The Sin of Science: Ignorance of Ignorance," Knowledge15, no. 2 (1993): 157-65.

48. Simon Jenkins in Times (London), December 6, 1997.

49. S. N. Cousens et al., "Sporadic Creutzfeldt-JakobDisease in the United Kingdom: Analysis of EpidemiologicalSurveillance Data for 1970-96," British Medical Journal 315(1997): 389-95.

50. John Collinge, "CJD Could Become an Epidemic ofBiblical Proportions," Times (London), August 7, 1997.

51. Christopher Booker in Sunday Telegraph, August 10,1997.

52. The National CJD Surveillance Unit, "Sixth AnnualReport 1997: Creutzfeldt-Jakob Disease Surveillance in theUK," http://www.cjd.ed.ac.uk/report97.html.

53. Almond and Pattison, p. 438.

54. From http://www.bse.org.uk.

55. D. H. Meadows, D. L. Meadows, and J. Randers, Beyondthe Limits: Global Collapse or a Sustainable Future(London: Earthscan, 1992), p. xv.

56. Julian Simon and Herman Kahn, eds., The ResourcefulEarth (Oxford: Blackwell, 1984), p. 3.

57. Martin Woollacott, "Risky Business, Safety" in ThePolitics of Risk Society, ed. Jane Franklin (Malden,Mass.: Blackwell, 1998), chapter 5.

58. Interview on "Big Science," conducted by David Malone,BBC2, August 22, 1995.

59. Quoted in the Times (London), February 8, 1997.

60. Julian Simon, The Ultimate Resource 2 (Princeton,N.J.: Princeton University Press, 1996), p. 15.

Published by the Cato Institute, Policy Analysis is a regular series evaluat-

ing government policies and offering proposals for reform. Nothing in Policy

Analysis should be construed as necessarily reflecting the views of the Cato

Institute or as an attempt to aid or hinder the passage of any bill before

congress. Contact the Cato Institute for reprint permission. Additional copies

of Policy Analysis are $6.00 each ($3.00 each for five or more). To order, or

for a complete listing of available studies, write the Cato Institute, 1000

Massachusetts Ave., N.W., Washington, D.C. 20001, call toll free 1-800-767-

1241 (noon - 9 p.m. eastern time), fax (202) 842-3490, or visit our website at

www.cato.org.

no. 335 march 4, 1999 cars, cholera, and cows · 2016-10-20 · no. 335 cars, cholera, and cows the...

Documents